Photon-counting CT: A Step Change Leading to a Revolution in Coronary Imaging

See also the article by Hagar et al in this issue.

Dr Michelle Williams is senior clinical lecturer and radiology consultant at the University of Edinburgh. She is associate director of the British Heart Foundation Data Science Centre. Her research centers around multimodality imaging of the heart, lungs, and blood vessels, including machine learning and advanced analytic techniques. She is president-elect of the British Society of Cardiovascular Imaging and a member of the Board of Directors of the Society of Cardiovascular Computed Tomography.

Professor David Newby is British Heart Foundation Duke of Edinburgh chair of cardiology at the University of Edinburgh, director of Edinburgh Imaging facilities, and a consultant interventional cardiologist at Royal Infirmary of Edinburgh. He has major interests in advanced imaging and clinical trials of cardiovascular disease and has been involved in several multicenter trials. He is chief investigator for the SCOT-HEART and SCOT-HEART 2 trials.

High resolution and excellent image quality are crucial in assessment of the coronary arteries on CT scans. Motion artifacts, contrast timing, image noise, body habitus, and calcification can all impact image quality and the diagnostic accuracy of coronary CT angiography (1). Spatial and temporal resolution are important factors that affect the ability to assess small structures, such as the coronary arteries, and even more so the assessment of atherosclerotic plaque characteristics. Photon-counting CT (PCCT) has recently emerged as a technique that is capable of superior high-resolution imaging and material decomposition analysis, which offers the promise of improved assessment of the coronary arteries and atherosclerotic plaque.

The arrival of PCCT has been much anticipated and, following the recent release of commercially available scanners, has generated major interest in a variety of clinical circumstances. Compared with a standard multidetector CT scanner that uses energy-integrated detectors, a PCCT scanner uses photon-counting detectors. These were originally developed for physics applications at the Large Hadron Collider at Conseil Européen pour la Recherche Nucléaire. Instead of converting photons to light, which is detected by photodiodes, photon-counting detectors directly count the number and energy of the photons that reach them. The use of PCCT scanners offers several potential advances for CT imaging. First, the ability to count photon numbers and energy means that multienergy spectral images can be reconstructed, with the potential to perform material decomposition analysis that could include better characterization of tissue components and multiagent contrast protocols. Second, photon-counting detectors may increase image resolution, leading to improved image quality, especially for small structures, such as coronary artery plaques. Third, because of improved image quality, there is the potential to develop new CT protocols with reduced radiation and contrast agent doses. However, at present, we do not know which patients would benefit most from this technology or how best to use it in clinical practice.

In this issue of Radiology, Hagar et al use (2) PCCT in ultrahigh-resolution mode to assess the severity of coronary artery disease in a prospective study of 68 individuals being considered for transcatheter aortic valve implantation and undergoing invasive coronary angiography. Their sample included individuals with high risk of coronary artery disease (CAD), with a mean age of 81 years and multiple cardiovascular risk factors. The area under the receiver operating characteristics curve of PCCT in the detection of CAD was 0.93 per participant, 0.94 per vessel, and 0.92 per segment. At the participant level, accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were 88%, 96%, 84%, 77%, and 97%, respectively, in the detection of CAD (≥50% stenosis) and 83%, 100%, 76%, 60%, and 100%, respectively, in the detection of obstructive CAD (≥70% stenosis). As expected, for stenosis of 50% or greater and for stenosis of 70% or greater, there was a decrease in per-vessel sensitivity (89% and 93%, respectively) and per-segment sensitivity (77% and 79%, respectively), along with a decrease in per-vessel positive predictive value (67% and 54%, respectively) and per-segment positive predictive value (43% and 30%, respectively). Overall, PCCT maintained high accuracy for the identification of stenosis of 50% or greater, even in study participants with a very high coronary artery calcium score (Agatston score ≥1000), known history of CAD, or prior stent implantation; however, accuracy in the identification of stenosis of 70% or greater was reduced in these subgroups. The authors estimate that when nondiagnostic segments were excluded from analysis, 54% of individuals in their study sample could have avoided invasive coronary angiography if management decisions were made based on coronary CT angiography with PCCT alone. This could benefit patients by providing a single imaging test for the simultaneous assessment of the coronary arteries and aortic valve, thereby reducing the need for unnecessary invasive investigations and helping diminish health care costs.

Coronary CT angiography with standard multidetector CT has good diagnostic accuracy in identification of the presence and severity of obstructive CAD in patients suspected of having CAD, with an excellent negative predictive value (1). This negative predictive value, the ability to rule out CAD or severe stenoses, is maintained even in patients with very high coronary artery calcium scores (>1000 Agatston units [AU]) (3). The diagnostic accuracy of coronary CT angiography in patients undergoing assessment for transcatheter aortic valve implantation is usually lower, due to the presence of more coronary artery calcification, greater likelihood of prior coronary intervention, and higher heart rates (4). The issue of coronary artery calcification remains a problem for PCCT, with the frequency of nondiagnostic segments increasing in patients with a calcium score greater than 600 AU (5). The coronary artery calcium scores in Hagar et al (2) were, on average, only moderately elevated, with a median score of 414 AU, and some individuals had no coronary artery calcification at all. While PCCT can improve diagnostic accuracy for coronary CT angiography, it does not remove this issue altogether, and extensive coronary calcification continues to cause problems of interpretation. Indeed, while reducing the need for invasive coronary angiography by 54% in the Hagar et al article (2) and by 40% in a previous meta-analysis (4) are welcome changes, these are arguably modest improvements and still leave nearly half of all patients requiring further potential imaging.

Coronary CT angiography has moved beyond the assessment of stenosis alone, and visual and quantitative analysis of atherosclerotic plaque burden and its subtypes are now an important component of image analysis and clinical risk stratification. How PCCT will impact plaque characterization and quantification, as well as radiomic analysis, remains to be established. It will be important to establish whether it can provide substantial advantages over and above those derived from analyses performed on standard CT angiography. In theory, there should be many potential advantages, but this will require new approaches, as well as major and substantial investments in computing power and analytical algorithms.

An important limitation of the protocol used in this study is the radiation dose. The mean radiation dose for the ultrahigh-resolution coronary CT angiography component, with retrospective electrocardiogram gating, was 13.3 mSv ± 4.2 (SD). This is high for coronary CT angiography which, with prospective electrocardiogram gating, we would expect to have a radiation dose of 2.9 mSv (6), or even below 1 mSv (7), using the same conversion factor. In addition, the conversion factor Hagar et al used for radiation dose calculation likely underestimates the radiation dose, as contemporary assessments have shown that conversion factors for coronary CT angiography are likely to be approximately double this (8). Although these radiation doses may be acceptable in older patients with high risk undergoing transcatheter aortic valve implantation or in the assessment of coronary artery stents, further hardware and software improvements to reduce radiation dose will be required for ultrahigh-resolution coronary CT angiography to be useful in broader clinical populations.

Ultrahigh-resolution imaging and multienergy imaging are not currently possible at the same time on this scanner. One of the most exciting applications of PCCT is the potential for multienergy imaging with material decomposition analysis and virtual noncontrast imaging, shown in a variety of phantom and animal studies. However, to date, in vivo clinical studies remain limited. Virtual noncontrast images can be useful in both coronary (9) and abdominal (10) imaging. We eagerly await further research to explore the accuracy and utility of material decomposition analysis.

Important questions remain about the application of this technology to clinical practice. How high a resolution is needed for coronary artery stenosis and plaque assessment? Do all patients require high-resolution imaging, or can we stratify its use, particularly if it is going to come at the cost of a higher radiation dose? How can we use this new information to guide care and improve outcomes for patients? There is no doubt that the introduction of PCCT has led to great excitement in the research and clinical communities. However, to translate this into improved outcomes for patients, these are important questions that must be addressed. In the meantime, ultrahigh-resolution coronary artery imaging is a step change for coronary CT angiography, but ultrahigh-resolution multienergy imaging may provide the revolution.